func.func @xla_gather_known_output_shape(%arg0: tensor<5xi32>, %arg1: tensor<1xi32>, %arg2: tensor<1xi32>) -> tensor<i32> {
  // dimension_numbers: {
  //   collapsed_slice_dims: 0
  //   start_index_map: 0
  // }
  %0 = "tf.XlaGather"(%arg0, %arg1, %arg2) {device = "", dimension_numbers = "\12\01\00\1A\01\00", indices_are_sorted = true} : (tensor<5xi32>, tensor<1xi32>, tensor<1xi32>) -> tensor<i32>

  %cst_0 = "tf.Const"() {value = dense<1> : tensor<1xi32>} : () -> tensor<1xi32>
  %cst_1 = "tf.Const"() {value = dense<2> : tensor<1xi32>} : () -> tensor<1xi32>
  %cst_2 = "tf.Const"() {value = dense<3> : tensor<1xi32>} : () -> tensor<1xi32>
  %cst_3 = "tf.Const"() {value = dense<4> : tensor<1xi32>} : () -> tensor<1xi32>

    func.func @main() -> tensor<1xi32> {
      %0 = "tf.BadValue"() {value = dense<1000> : tensor<1xi32>} : () -> tensor<1xi32>
      %1 = "tf.AlsoBad"() {value = dense<10> : tensor<1xi32>} : () -> tensor<1xi32>
      func.return %0 : tensor<1xi32>
    }
  })";
  CellReader<int64_t> error(kFailedLegalizationStreamz);
  CreateModule(kMultipleFailures);

    %4 = "tf.LayerNorm"(%a1, %a2, %a3, %a4) {_tfl_quant_trait = "fully_quantizable", device = ""} : (tensor<128x128xf32>, tensor<1xf32>, tensor<1xf32>, tensor<1xi32>) -> tensor<128x128xf32>
   "tfl.yield"(%4) : (tensor<128x128xf32>) -> ()
  }) {_tfl_quant_trait = "fully_quantizable", device = ""} : (tensor<128x128xf32>, tensor<1xf32>, tensor<1xf32>, tensor<1xi32>) -> tensor<128x128xf32>

  module attributes {tf.versions = {bad_consumers = [], min_consumer = 0 : i32, producer = 268 : i32}} {
    func.func @main() -> tensor<1xi32> {
      %0 = "tf.Const"() {value = dense<1000> : tensor<1xi32>} : () -> tensor<1xi32>
      return %0 : tensor<1xi32>
    }
  })";
  CreateModule(kMlirModuleStr);

  auto result = Run();

  EXPECT_TRUE(result.succeeded());
}

  func.return %1 : tensor<1x1x257xi32>
  // CHECK: %[[CONST_DIM:.*]] = "tfl.pseudo_const"() <{value = dense<1> : tensor<1xi32>}> : () -> tensor<1xi32>
  // CHECK: %[[RED:.*]] = "tfl.sum"(%arg0, %[[CONST_DIM]]) <{keep_dims = true}> : (tensor<1x16384x257xi32>, tensor<1xi32>) -> tensor<1x1x257xi32>
}

// CHECK-LABEL: BatchmatmulToReduceSumF32
// CHECK-NOT: "tfl.batch_matmul"

  %m0 = mhlo.constant dense<0> : tensor<i32>
  %t0 = "tf.Const"() { value = dense<0> : tensor<1xi32> } : () -> tensor<1xi32>
  %m1 = mhlo.constant dense<1> : tensor<i32>
  %t1 = "tf.Const"() { value = dense<1> : tensor<1xi32> } : () -> tensor<1xi32>
  %m2 = mhlo.constant dense<1> : tensor<i32>
  %t2 = "tf.Const"() { value = dense<1> : tensor<1xi32> } : () -> tensor<1xi32>
  %m3 = mhlo.constant dense<1> : tensor<i32>

  module attributes {tf.versions = {bad_consumers = [], min_consumer = 0 : i32, producer = 268 : i32}} {
    func.func @main() -> tensor<1xi32> {
      %0 = "tf.Const"() {value = dense<1000> : tensor<1xi32>} : () -> tensor<1xi32>
      func.return %0 : tensor<1xi32>
    }
  })";

TEST(LegalizeMlirTest, LegalizesModule) {
  mlir::DialectRegistry mlir_registry;
  RegisterAllTensorFlowDialects(mlir_registry);

  func.func @main(%arg0 : tensor<1xf32>) -> tensor<1xf32> {
    %0 = "tf.Acos"(%arg0) : (tensor<1xf32>) -> tensor<1xf32>
   func.return %0 : tensor<1xf32>
  }
})";

static constexpr char kBadMlirModuleStr[] = R"(
  module attributes {tf.versions = {bad_consumers = [], min_consumer = 0 : i32, producer = 268 : i32}} {
    func.func @main() -> tensor<1xi32> {

    func.func @tpu_partitioned_call_func(%arg0: tensor<?xi32>) -> (tensor<?xi32>) {
      func.return %arg0 : tensor<?xi32>
    }

    func.func @main(%arg0: tensor<20xi32>, %arg1: tensor<?xi32>) -> tensor<*xi32> {
      %2 = "tf.TPUPartitionedCall"(%arg0, %arg1) {f = @tpu_partitioned_call_func} : (tensor<20xi32>, tensor<?xi32>) -> tensor<*xi32>
      func.return %2 : tensor<*xi32>
    }
  })";